The IEEE Computer Society (CS) and Computer Vision Foundation (CVF) have reported that the 2025 Conference on Computer Vision and Pattern Recognition (CVPR) closed with “record-breaking technical success.”
With the highest number of paper submissions ever—12,008, representing a 13 percent increase over 2024—and 9,375 registrants from 75 different countries and regions, CVPR created a vibrant climate for discovery, collaboration, and advancement in the field of computer vision.
“CVPR continues to set the bar for research in computer vision,” stated Bryan Morse, CVPR 2025 General Co-Chair and Professor of Computer Science at Brigham Young University in Provo, Utah, US. “We can expect the research from this year’s event to be cited into the future.”
Technical Program
This year’s conference took place from June 11-15 in Nashville, Tennessee, drawing together the global computer vision community to share the latest advancements in the field. Major theme areas included image and video synthesis, 3D modeling from multi-view and sensors, multi-modal learning, and human-centric topics like face and body analysis.
Given that CVPR proceedings earned the number two spot in Google’s 2024 Scholar Metrics, the high caliber of accepted papers is likely to continue elevating the conference’s reputation. This year, only 22.1 percent of submitted papers were accepted for presentation, and just 3.3 percent were selected for oral presentations. Such rigorous standards emphasize the competitive nature of the conference.
Additionally, the 118 workshops and 25 tutorials offered thorough insights into prominent use cases for computer vision, spanning applications from autonomous vehicles to medicine and robotics. “As the field advances, we are seeing more and more applications of computer vision across various disciplines,” remarked Phillip Isola, CVPR 2025 program co-chair and an associate professor at the Massachusetts Institute of Technology (MIT) in Boston, Massachusetts, US.
Award-Winning Papers
Exciting advancements were highlighted by the best papers of the conference, which are poised to shape future research directions.
- VGGT: Visual Geometry Grounded Transformer was named Best Paper for introducing a feed-forward neural network that estimates key 3D scene properties for hundreds of input views, surpassing traditional approaches.
- Neural Inverse Rendering from Propagating Light, recognized as Best Student Paper, demonstrated a novel method for modeling and reverting multi-viewpoint, time-resolved measurements of propagating light from a LiDAR system, recovering scene geometry and rendering videos of light dynamics.
Fuxin Li, CVPR 2025 program co-chair and associate professor at Oregon State University in Corvallis, Oregon, US, stated, “All of the papers accepted to CVPR help to advance the work of the community. It will be exciting to watch how research continues to unfold over the next year.”
Keynote Presentations
CVPR 2025 featured keynotes that addressed transformative topics such as how artificial intelligence (AI) can bolster the low-altitude economy and enhance robotics through advanced AI integration. Notable speakers included:
- Harry Shum, former executive vice president of Microsoft Corporation and current Council Chairman of Hong Kong University of Science and Technology, presenting “Exploring the Low Altitude Airspace: From Natural Resource to Economic Engine.”
- Laurens van der Maaten, Distinguished Research Scientist at Meta AI, with “The Llama Herd of Models: System 1, 2, 3 Go!”
- Carolina Parada, Senior Director and Head of Robotics at Google DeepMind, discussing “Gemini Robotics: Bringing AI to the Physical World.”
Expo Highlights and Demo Awards
The conference showcased nearly 100 leading computer vision companies from around the world, reflecting the growing interest in applied solutions. This year’s expo included 69 demos, a significant 33 percent increase from 2024, displaying innovative commercial research and applications.
Among the top demos recognized by the organizers were:
- Best Demo: DynaMem and Robot Utility Models; contributed by Haritheja Etukuru, Peiqi Liu, Chris Paxton, Soumith Chintala, Lerrel Pinto, Nur Muhammad Mahi Shafiullah.
- Honorable Mention: MASt3R-SLAM; developed by Riku Murai, Eric Dexheimer, Andrew J. Davison.
- Honorable Mention: UltraFusion HDR; by Yujin Wang, Zixuan Chen, Xin Cai, Zhiyuan You, Zheming Lu, Fan Zhang, Shi Guo, Tianfan Xue.
- Honorable Mention: AI3D Render; created by Yosun Chang and Ryan Burgert.
Prominent sponsors of the conference, who unveiled new products and research, included Adobe, Amazon Science, AMD, Apple, ByteDance, GMI, Google, Intel, Meta, Qualcomm, Sony, and Tencent. A complete list of sponsors can be found on the CVPR website.
AI Art Awards
At the intersection of creativity and technology, the CVPR AI Art Gallery showcased 102 works by artists and researchers. Out of these, 73 pieces were exhibited both in person and online, while 29 were featured exclusively online. The top three prize winners were:
- Tom White for Atlas of Perception.
- Masaru Mizuochi for Green Diffusion.
- Mingyong Cheng, Sophia Sun, and Han Zhang for Learning to Move, Learning to Play, Learning to Animate.
Broadening Participation
The IEEE CS, CVF, and the Program Committee demonstrated a commitment to inclusivity by awarding $200,000 in travel grants and waiving registration fees for 276 deserving applicants from around the globe, ensuring that a diverse range of voices and talents were represented at the conference.
Looking Ahead
As new applications of computer vision and AI continue to evolve, the work showcased at CVPR will be essential in forging the future of these technologies. Keep an eye out for ongoing advancements from IEEE CS and CVF throughout the year, and mark your calendars for CVPR 2026, taking place from June 6-12 in Denver, Colorado, US.